On the algorithmic censorship of images
Contemporary urban culture is inseparable from image production on networks. Never have so many photographs been taken as in our time. By the end of 2017, 1.2 trillion images are expected to be shared, an increase of 9% over last year. An estimated 3.5 trillion film photographs have been produced since photography’s invention in the 19th century. In other words, in the past two years, an amount of images has flowed through networks that humanity took almost two centuries to produce by analog means.
The change is not merely quantitative. It is qualitative too. It involves an unprecedented process in which new social subjects appropriate the screen and the image frame. From their beginnings, images were directly associated with class, gender and political power, reserved first for sacred figures, kings, aristocrats and popes, and later for politicians and the wealthy bourgeoisie. Throughout the 20th century, mass communications broadened the range of people who could become a published image, potentially even an archived one. But only in the 21st century, with digital cameras and the internet, can we speak of the large-scale multiplication and diversification of the social and cultural spectrum of visual records.
The protagonist of this story is the camera-equipped, internet-connected mobile phone. It transformed the camera from a recording device into a device for projecting the self. That personal projection has a definite destination: social networks and interpersonal groups, from Facebook and Instagram to Snapchat and WhatsApp. In these circuits, images are linked to the place and time of their production, contextualize each author in relation to their group, and are tracked according to certain patterns. This is where the culture of sharing intersects with the culture of surveillance.
This ambivalence is one of the defining traits of network culture and points to different political conceptions and tendencies in today’s media ecology. Networks open unprecedented possibilities for the critical and creative use of existing media, but they are also new platforms for encouraging consumption and controlling people, who become easily traceable through devices of every kind. This tracking is performed not only by urban infrastructure equipment; it is fueled by the wealth of data we voluntarily provide, especially through our use of mobile phone cameras.
Unlike texts, which can be tracked semantically, visual content is mapped by recognizing patterns such as lines, densities and geometric forms. These patterns designate, for example, what are supposedly breasts, buttocks and penises. They can therefore act as the first agents of image censorship on social networks, something that is becoming increasingly commonplace. Facebook and its affiliated company Instagram do not disclose whether they use algorithms to screen material that is blocked. But given the information overload circulating in these spaces, it is quite plausible that the process follows, or will soon follow, the same logic as the automated management of Facebook’s timeline, which “chooses” the content we see through algorithmic instructions.
Image blocking affects everything from historical content to contemporary art and influences how we decide to use networks. One example is Walter Grabe’s 1909 photograph of the Botocudo Indigenous people, which led to the blocking of a Ministry of Culture Facebook page. It shows an Indigenous woman with bare breasts and had been chosen to illustrate the 2015 launch of the Brasiliana Fotográfica portal. After protests, including from the Brazilian government, it was released. Other, more recent cases, such as “genderless nipples,” reveal more political nuances in selection processes.
In a world increasingly mediated by visual culture and its prohibitions, pioneering projects such as artist Christophe Bruno’s Logo Hallucination (2006) gain relevance. In a humorous act of peaceful guerrilla warfare, Bruno selected 18 major corporate logos, including McDonald’s, AT&T, Apple and Mercedes-Benz, and set pattern-recognition software to work searching the internet for images that might be considered the originals of those corporations’ logos.
The result is hallucinatory, to say the least, as the project’s name suggests: the Atari logo was supposedly already present in a Vermeer painting, an African mask was the original of the McDonald’s logo, among other bizarre examples documented on the project’s website. Bruno showed how new pattern-recognition tools were becoming fertile ground for technologies to control and manage image copyrights. He asked whether we might reach such a degree of hallucination that it would culminate in the privatization of the gaze. Viewed ten years later, Logo Hallucination emerges as an introductory lesson in the methods of a new form of censorship. A censorship that does not prohibit. Instead, it algorithmically defines the right to see.
Originally published in Dissenso.org